Engineered proteins, such as multispecific antibodies that bind two or more antigens are known in the art. Such multispecific binding proteins can be generated using cell fusion, chemical conjugation, or recombinant DNA techniques.
Bispecific antibodies have been produced using quadroma technology (see Milstein and Cuello (1983) Nature 305(5934):537-40) based on the somatic fusion of two different hybridoma cell lines expressing murine monoclonal antibodies (mAbs) with the desired specificities of the bispecific antibody. Because of the random pairing of two different immunoglobulin (Ig) heavy and light chains within the resulting hybrid-hybridoma (or quadroma) cell line, up to ten different Ig species are generated, of which only one is the functional bispecific antibody. The presence of mis-paired by-products, and significantly reduced production yields, means sophisticated purification procedures are required.
Bispecific antibodies can also be produced by chemical conjugation of two different mAbs (see Staerz et al. (1985) Nature 314(6012):628-31). This approach does not yield homogeneous preparation. Other approaches have used chemical conjugation of two different mAbs or smaller antibody fragments (see Brennan et al. (1985) Science 229(4708):81-3).
Another method used to produce bispecific antibodies is the coupling of two parental antibodies with a hetero-bifunctional crosslinker, but the resulting bispecific antibodies suffer from significant molecular heterogeneity because reaction of the crosslinker with the parental antibodies is not site-directed. To obtain more homogeneous preparations of bispecific antibodies two different Fab fragments have been chemically crosslinked at their hinge cysteine residues in a site-directed manner (see Glennie et al. (1987) J. Immunol. 139(7):2367-75). But this method results in Fab′2 fragments, not full IgG molecule.
A wide variety of other recombinant bispecific antibody formats have been developed (see Kriangkum et al. (2001) Biomol. Eng. 18(2):31-40). Amongst them tandem single-chain Fv molecules and diabodies, and various derivatives thereof, are the most widely used. Routinely, construction of these molecules starts from two single-chain Fv (scFv) fragments that recognize different antigens (see Economides et al. (2003) Nat. Med. 9(1):47-52). Tandem scFv molecules (taFv) represent a straightforward format simply connecting the two scFv molecules with an additional peptide linker. The two scFv fragments present in these tandem scFv molecules form separate folding entities. Various linkers can be used to connect the two scFv fragments and linkers with a length of up to 63 residues (see Nakanishi et al. (2001) Ann. Rev. Immunol. 19:423-74). Although the parental scFv fragments can normally be expressed in soluble form in bacteria, it is, however, often observed that tandem scFv molecules form insoluble aggregates in bacteria. Hence, refolding protocols or the use of mammalian expression systems are routinely applied to produce soluble tandem scFv molecules. In a recent study, in vivo expression by transgenic rabbits and cattle of a tandem scFv directed against CD28 and a melanoma-associated proteoglycan was reported (see Gracie et al. (1999) J. Clin. Invest. 104(10):1393-401). In this construct, the two scFv molecules were connected by a CH1 linker and serum concentrations of up to 100 mg/L of the bispecific antibody were found. Various strategies including variations of the domain order or using middle linkers with varying length or flexibility were employed to allow soluble expression in bacteria. A few studies have now reported expression of soluble tandem scFv molecules in bacteria (see Leung et al. (2000) J. Immunol. 164(12):6495-502; Ito et al. (2003) J. Immunol. 170(9):4802-9; Karni et al. (2002) J. Neuroimmunol. 125(1-2):134-40) using either a very short Ala3 linker or long glycine/serine-rich linkers. In a recent study, phage display of a tandem scFv repertoire containing randomized middle linkers with a length of 3 or 6 residues was employed to enrich for those molecules that are produced in soluble and active form in bacteria. This approach resulted in the isolation of a tandem scFv molecule with a 6 amino acid residue linker (see Arndt and Krauss (2003) Methods Mol. Biol. 207:305-21). It is unclear whether this linker sequence represents a general solution to the soluble expression of tandem scFv molecules. Nevertheless, this study demonstrated that phage display of tandem scFv molecules in combination with directed mutagenesis is a powerful tool to enrich for these molecules, which can be expressed in bacteria in an active form.
Bispecific diabodies (Db) utilize the diabody format for expression. Diabodies are produced from scFv fragments by reducing the length of the linker connecting the VH and VL domain to approximately 5 residues (see Peipp and Valerius (2002) Biochem. Soc. Trans. 30(4):507-11). This reduction of linker size facilitates dimerization of two polypeptide chains by crossover pairing of the VH and VL domains. Bispecific diabodies are produced by expressing, two polypeptide chains with, either the structure VHA-VLB and VHB-VLA (VH-VL configuration), or VLA-VHB and VLB-VHA (VL-VH configuration) within the same cell. A large variety of different bispecific diabodies have been produced in the past and most of them are expressed in soluble form in bacteria. However, a recent comparative study demonstrates that the orientation of the variable domains can influence expression and formation of active binding sites (see Mack et al. (1995) Proc. Natl. Acad. Sci. USA 92(15):7021-5). Nevertheless, soluble expression in bacteria represents an important advantage over tandem scFv molecules. However, since two different polypeptide chains are expressed within a single cell inactive homodimers can be produced together with active heterodimers. This necessitates the implementation of additional purification steps in order to obtain homogenous preparations of bispecific diabodies. One approach to force the generation of bispecific diabodies is the production of knob-into-hole diabodies (see Holliger et al. (1993) Proc. Natl. Acad. Sci. USA 90(14):6444-8.18). This was demonstrated for a bispecific diabody directed against HER2 and CD3. A large knob was introduced in the VH domain by exchanging Val37 with Phe and Leu45 with Trp and a complementary hole was produced in the VL domain by mutating Phe98 to Met and Tyr87 to Ala, either in the anti-HER2 or the anti-CD3 variable domains. By using this approach the production of bispecific diabodies could be increased from 72% by the parental diabody to over 90% by the knob-into-hole diabody. Importantly, production yields only slightly decrease as a result of these mutations. However, a reduction in antigen-binding activity was observed for several constructs. Thus, this rather elaborate approach requires the analysis of various constructs in order to identify those mutations that produce heterodimeric molecule with unaltered binding activity. In addition, such approach requires mutational modification of the immunoglobulin sequence at the constant region, thus creating non-native and non-natural form of the antibody sequence, which may result in increased immunogenicity, poor in vivo stability, as well as undesirable pharmacokinetics.
Single-chain diabodies (scDb) represent an alternative strategy for improving the formation of bispecific diabody-like molecules (see Holliger and Winter (1997) Cancer Immunol. Immunother. 45(3-4):128-30; Wu et al. (1996) Immunotechnology 2(1):21-36). Bispecific single-chain diabodies are produced by connecting the two diabody-forming polypeptide chains with an additional middle linker with a length of approximately 15 amino acid residues. Consequently, all molecules with a molecular weight corresponding to monomeric single-chain diabodies (50-60 kDa) are bispecific. Several studies have demonstrated that bispecific single chain diabodies are expressed in bacteria in soluble and active form with the majority of purified molecules present as monomers (see Holliger and Winter (1997) Cancer Immunol. Immunother. 45(3-4):128-30; Wu et al. (1996) Immunotechnol. 2(1):21-36; Pluckthun and Pack (1997) Immunotechnol. 3(2):83-105; Ridgway et al. (1996) Protein Engin. 9(7):617-21). Thus, single-chain diabodies combine the advantages of tandem scFvs (all monomers are bispecific) and diabodies (soluble expression in bacteria).
More recently diabodies have been fused to Fc to generate more Ig-like molecules, named di-diabodies (see Lu et al. (2004) J. Biol. Chem. 279(4):2856-65). In addition, multivalent antibody constructs comprising two Fab repeats in the heavy chain of an IgG and that bind four antigen molecules have been described (see PCT Publication No. WO 0177342, and Miller et al. (2003) J. Immunol. 170(9):4854-61).
There is a need in the art for improved multivalent binding proteins that bind two or more antigens. U.S. Pat. No. 7,612,181 provides a novel family of binding proteins that bind two or more antigens with high affinity, and which are called dual variable domain immunoglobulins (DVD-Ig™). The present invention provides further novel binding proteins that bind two or more antigens.